EpiTrace.AI
Science

How cells write their history into DNA

EpiTrace.AI rests on two decades of research into DNA methylation. This page explains the biology, the measurement methods and the analysis steps in plain language, shows where the platform fits and lists the studies worth reading.

01 · Basics

DNA methylation in brief

Every cell in the body carries the same DNA, yet a liver cell behaves differently from a blood cell. One reason is DNA methylation: small chemical tags, methyl groups, attached to the DNA letter C where it is followed by G. Researchers call these places CpG sites [26].

Methylation helps to switch genes on or off and keeps a cell's identity stable through many divisions. It is measured per site as a value between 0 and 1, the share of DNA copies that carry the tag. Scientists call this the β-value.

Patterns change during development, with age and in disease. In cancer, whole regions gain or lose methylation, and those changes are passed on when tumour cells divide [1].

cell_07ancestor
cell_31daughter
cell_58granddaughter · new branch
β-value: share of methylated copies at one CpG site
02 · Measurement

Measuring methylation in single cells

Measurements across millions of cells give an average. Single-cell methods keep the differences between cells, the raw material for lineage tracing. Each method trades the number of CpG sites per cell against the number of cells [3][4].

Genome-wide, bisulfite-based

scBS-seq, scRRBS and snmC-seq read methylation across the genome. Bisulfite turns unmethylated C into U, so sequencing shows which sites were methylated. Deep per cell, fewer cells per experiment [6][7][8].

Targeted, enzyme-based

scTAM-seq measures hundreds of selected CpG sites in thousands of cells. Restriction enzymes cut only unmethylated sites, so no bisulfite is needed. Fewer sites, many more cells [11].

Several layers per cell

scM&T-seq and scNMT-seq add gene activity and open chromatin to methylation from the same cell. This links marks to their effect [9][10].

Sites per cell versus cells per experimentcells per experiment →CpG sites per cell →scBS-seqscRRBSsnmC-seqscM&T-seqscNMT-seqscTAM-seqBisulfiteMulti-omicsTargeted
Qualitative positioning based on the method papers [6]–[11]; exact numbers depend on protocol and depth.
03 · Lineage tracing

From inherited marks to family trees

When a cell divides, a few methylation marks change at random. The daughter cells inherit these changes, and so do their daughters. Over many divisions, each branch of a cell family collects its own combination, much like a family name passed down the generations.

Reading these combinations lets researchers reconstruct which cells share an ancestor without adding genetic barcodes. Independent studies have used this principle to trace the history of leukaemia cells in patients [13] and the clonal dynamics of blood formation over a lifetime [14]. A review sets out the opportunities and limits of combining lineage tracing with single-cell measurement [5].

Two kinds of sites matter. Sites that change with cell type show what a cell is. Sites that change rarely and at random show where it came from. Good analyses keep both apart.

  1. Ancestorstarting pattern
  2. Daughtersinherit the pattern
  3. New brancha random change marks a sub-family
  4. Cloneall cells that share the change
04 · Analysis chain

From the lab to an insight, step by step

EpiTrace.AI covers the middle of the chain, where data becomes understandable. Lab work and validation stay with the research groups.

  1. 01 · Lab and pipelines

    Measure

    The lab produces raw single-cell reads.

  2. 02 · Lab and pipelines

    Process

    Pipelines such as Bismark align reads and call methylation per CpG site.

  3. 03 · EpiTrace.AI today

    Load

    A matrix of cells × CpG sites with annotations, stored as .h5ad or CSV.

  4. 04 · EpiTrace.AI today

    Map

    Dimension reduction and UMAP place similar cells close together [22].

  5. 05 · EpiTrace.AI today

    Group and follow

    Clusters and lineages are shown side by side; reconstruction directly from inherited changes follows in v2X.

  6. 06 · EpiTrace.AI today

    Rank markers

    CpG sites are ranked by how clearly they separate groups.

  7. 07 · v2X and lab

    Explain and validate

    SHAP values show why [24]; lab experiments confirm or reject candidates.

Lab and pipelinesEpiTrace.AI todayv2X and lab
05 · Ageing

Epigenetic clocks: methylation as a measure of age

Some CpG sites change so regularly with age that a few hundred of them together estimate the age of a tissue. The first multi-tissue clock appeared in 2013 [15]. Since then, researchers have shown that epigenetic age can be estimated even in single cells [16].

For EpiTrace.AI this matters in two ways: age-related sites must be separated from lineage signals, and age itself can become a layer on the cell map in later versions.

06 · Open questions

What the field is still working on

Sparse data

Single cells yield only part of their methylome. Statistical models and imputation fill gaps, with care [17][18].

Batch effects

Differences between runs can look like biology. Shared standards and documented processing steps make results comparable.

Ground truth

Reconstructed family trees need checks against known lineages, for example genetic barcodes in model systems [14].

Data protection

Methylation data can be personal data. Research needs clear data rules, consent and storage in the EU.

07 · Reading list

Studies and articles worth reading

26 sources, checked on 19 November 2026. They are independent research by other groups and are listed as background, not as results of EpiTrace.AI.

Foundations and reviews

  1. [1]
    Jones PA, Baylin SB (2007). The epigenomics of cancer. Cell 128:683–692.Classic overview of how epigenetic changes contribute to cancer.doi.org/10.1016/j.cell.2007.01.029
  2. [2]
    Roadmap Epigenomics Consortium, Kundaje A et al. (2015). Integrative analysis of 111 reference human epigenomes. Nature 518:317–330.Reference epigenomes of human tissues and cell types; a common baseline for comparisons.doi.org/10.1038/nature14248
  3. [3]
    Karemaker ID, Vermeulen M (2018). Single-cell DNA methylation profiling: technologies and biological applications. Trends in Biotechnology 36:952–965.Readable review of single-cell methylation methods and what they reveal.doi.org/10.1016/j.tibtech.2018.04.002
  4. [4]
    Liu F et al. (2023). Technologies and applications of single-cell DNA methylation sequencing. Theranostics 13:2439–2454.More recent review of methods and applications, including cancer.doi.org/10.7150/thno.82582
  5. [5]
    Wagner DE, Klein AM (2020). Lineage tracing meets single-cell omics: opportunities and challenges. Nature Reviews Genetics 21:410–427.How lineage tracing and single-cell measurements come together, and where the limits are.doi.org/10.1038/s41576-020-0223-2

Single-cell methods

  1. [6]
    Smallwood SA et al. (2014). Single-cell genome-wide bisulfite sequencing for assessing epigenetic heterogeneity. Nature Methods 11:817–820.scBS-seq: genome-wide methylation of individual cells.doi.org/10.1038/nmeth.3035
  2. [7]
    Guo H et al. (2013). Single-cell methylome landscapes of mouse embryonic stem cells and early embryos analyzed using reduced representation bisulfite sequencing. Genome Research 23:2126–2135.scRRBS: methylation of CpG-rich regions in single cells.genome.cshlp.org/content/23/12/2126
  3. [8]
    Luo C et al. (2017). Single-cell methylomes identify neuronal subtypes and regulatory elements in mammalian cortex. Science 357:600–604.snmC-seq: methylomes of thousands of single nuclei from tissue.doi.org/10.1126/science.aan3351
  4. [9]
    Angermueller C et al. (2016). Parallel single-cell sequencing links transcriptional and epigenetic heterogeneity. Nature Methods 13:229–232.scM&T-seq: methylation and gene activity from the same cell.doi.org/10.1038/nmeth.3728
  5. [10]
    Clark SJ et al. (2018). scNMT-seq enables joint profiling of chromatin accessibility, DNA methylation and transcription in single cells. Nature Communications 9:781.scNMT-seq: three layers from one cell, a model for multi-omics.doi.org/10.1038/s41467-018-03149-4
  6. [11]
    Bianchi A, Scherer M et al. (2022). scTAM-seq enables targeted high-confidence analysis of DNA methylation in single cells. Genome Biology 23:229.scTAM-seq: hundreds of selected CpG sites in thousands of cells, without bisulfite.doi.org/10.1186/s13059-022-02796-7
  7. [12]
    Vaisvila R et al. (2021). Enzymatic methyl sequencing detects DNA methylation at single-base resolution from picograms of DNA. Genome Research 31:1280–1289.EM-seq: an enzymatic alternative to bisulfite conversion for very small amounts of DNA.genome.cshlp.org/content/31/7/1280

Lineage tracing, cancer and ageing

  1. [13]
    Gaiti F et al. (2019). Epigenetic evolution and lineage histories of chronic lymphocytic leukaemia. Nature 569:576–580.Single-cell methylation used to reconstruct lineage histories in a human cancer.doi.org/10.1038/s41586-019-1198-z
  2. [14]
    Scherer M et al. (2025). Clonal tracing with somatic epimutations reveals dynamics of blood ageing. Nature 643:478–487.Clonal tracing from naturally occurring methylation changes, without genetic barcodes.doi.org/10.1038/s41586-025-09041-8
  3. [15]
    Horvath S (2013). DNA methylation age of human tissues and cell types. Genome Biology 14:R115.The multi-tissue epigenetic clock that started the field of methylation age.doi.org/10.1186/gb-2013-14-10-r115
  4. [16]
    Trapp A, Kerepesi C, Gladyshev VN (2021). Profiling epigenetic age in single cells. Nature Aging 1:1189–1201.Epigenetic age estimated in individual cells.doi.org/10.1038/s43587-021-00134-3

Computation and AI

  1. [17]
    Kapourani CA et al. (2021). scMET: Bayesian modeling of DNA methylation heterogeneity at single-cell resolution. Genome Biology 22:114.Statistical model for variability between cells, sparse data included.doi.org/10.1186/s13059-021-02329-8
  2. [18]
    Angermueller C et al. (2017). DeepCpG: accurate prediction of single-cell DNA methylation states using deep learning. Genome Biology 18:67.Deep learning that fills gaps in single-cell methylation data.doi.org/10.1186/s13059-017-1189-z
  3. [19]
    Kremer LPM et al. (2024). Analyzing single-cell bisulfite sequencing data with MethSCAn. Nature Methods.Practical workflow for exploring single-cell methylomes.doi.org/10.1038/s41592-024-02347-x
  4. [20]
    Wolf FA, Angerer P, Theis FJ (2018). SCANPY: large-scale single-cell gene expression data analysis. Genome Biology 19:15.The Python toolkit EpiTrace.AI builds on.doi.org/10.1186/s13059-017-1382-0
  5. [21]
    Virshup I et al. (2024). anndata: Access and store annotated data matrices. Journal of Open Source Software.The data format behind .h5ad files, which EpiTrace.AI reads.doi.org/10.21105/joss.04371
  6. [22]
    McInnes L, Healy J, Melville J (2018). UMAP: Uniform Manifold Approximation and Projection for dimension reduction. arXiv:1802.03426.The method behind the cell maps.arxiv.org/abs/1802.03426
  7. [23]
    Lange M et al. (2022). CellRank for directed single-cell fate mapping. Nature Methods 19:159–170.Estimating where cells are heading; a reference for planned fate estimates.doi.org/10.1038/s41592-021-01346-6
  8. [24]
    Lundberg SM, Lee S-I (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems 30.SHAP values, the basis of the planned explanation layer.arxiv.org/abs/1705.07874

Data stewardship and introductions

  1. [25]
    Wilkinson MD et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data 3:160018.Findable, accessible, interoperable, reusable: the principles the project follows.doi.org/10.1038/sdata.2016.18
  2. [26]
    National Human Genome Research Institute. Epigenomics Fact Sheet.Short, accessible introduction for non-specialists.www.genome.gov/about-genomics/fact-sheets/Epigenomics-Fact-Sheet